ANNOUNCE: A *second* Python 2.1 release candidate!

[Tim Peters]

[Don Tuttle]
> ...
> Tim, you're the one who taught me what a pain in the ass it is to
> release a new version of Python.

The pain of making a release is proportional to some power of the number of
changes made since the last release, and that means changes in both the
Python codebase and in the platforms people are running on.  Between 2.1c1
and 2.1c2, only a handful of changes were made to the Python codebase, and
AFAIK nobody popped up with a popular new incompatible flavor of Unix between
Friday and Sunday <wink>.  So releases are not *uniformly* painful.  Cutting
these rapid releases at the end has been easy.

> If you and the dev team felt that 2.1 was ready to be released last
> Friday (the 13th), it would have been fine with me.  But you didn't.
> Instead you went to the trouble to build another release.

Actually three times since Friday:  2.1c1 then, 2.1c2 Sunday, and will be 2.1
Tuesday morning.  Since all the *hard* work has been done, and the rate of
change has slowed to a drip, we could cut half a dozen new release candidates
every day now, if that seemed of use (but it doesn't -- if you're complaining
that 4 days isn't enough to download it and try it, I can't imagine you'd be
happier with 4 hours <wink>).

> It's the fact that you went to all this trouble,

The 2.1c series has been easy.

> but only allowed 4 days of use, that brought me into this discussion.
> This seems rushed.  Why go to all this trouble if you're not going to
> give users the time to test on all these different platforms?

I've already answered this:  people have been testing for months.  There's no
rush.  Neither is there any point to extending the prerelease period beyond
its usefulness.  The people keen to test are already testing, and others
can't be persuaded to try -- waiting even months won't change that, and
delaying releases artificially is as silly as pushing them out early.  The
major point of the rc series was to get a close-to-2.1-final-as-possible
prerelease package into the hands of people who reported new glitches over
the last few weeks, to make sure the fixes really fix their problems on their
platforms, and don't create gross new problems for others.  The former people
certainly are trying the packages, and they're primarily who we made the
packages for.  That at least hundreds of other people give them a try too is
gravy.

>...
> It seems this is being done simply to meet some arbitrary
> deadline

It's being done because the release is ready on technical grounds.  The
release candidates aren't a sign of haste on Guido's part, but of caution:
asking people to test fixes from a prerelease package is more reliable than
asking people to test from their own private copy of the CVS tree, and also
much more convenient for them (which greatly increases our chances that they
*will* try it).

Alas, history tells us there's no way to get *huge* numbers before a final
release is out the door.  In that sense, and like it or not, putting out 2.1
final is the only way we have to get the only kind of broad 2.1 field testing
you seem to think is worth the effort (and on that we definitely disagree --
rc1 and rc2 were certainly worth the (marginal!) effort of putting together,
rc1 because people did report last-second brainos in it, and rc2 because they
haven't reported more).

Now my best guess is that within 3 years, we'll have millions of people
paying us $30 each just to get their hands on a Python alpha release.  It
works for Bill Gates, and Guido's smarter than him <wink>.

until-then-we-gratefully-take-whatever-people-freely-give-but-
    can't-get-more-than-that-ly y'rs  - tim